Summertime Convective Initiation Nowcasting over Southeastern China 1 Based on Advanced Himawari Imager Observations

12 Convective initiation (CI) nowcasting often has a low probability of detection (POD) and 13 a high false-alarm ratio (FAR) at sub-tropical regions where the warm-rain processes often 14 occur. Using the high spatial- and temporal-resolution and multi-spectral data from the 15 Advanced Himawari Imager (AHI) on board Japanese new-generation geostationary satellite 16 Himawari-8, a stand-alone CI nowcasting algorithm is developed in this study. The 17 AHI-based CI algorithm utilizes the reflectance observations from channels 1 (0.47 μm ) and 7 18 (3.9 μm ), brightness temperature observations from infrared window channel 13 (10.4 μm ), the 19 dual-spectral differences between channels 10 (7.3 μm ) and 13, 13 and 15 (12.4 μm ), as well as 20 a tri-spectral combination of channels 11, 15 and 13, as CI predictors without relying on any 21 dynamic ancillary data (e.g., cloud type and atmospheric motion vector products). The 22 proposed AHI-based algorithm is applied to CI cases over Fujian province in the 23 Southeastern China. When validated by S-band radar observations, the CI algorithm 24 produced a POD as high as 93.33%, and a FAR as low as 33.33% for a CI case day that 25 occurred on 1 August 2015 over Northern Fujian. For over 216 CI events that occurred in a 26 three-month period from July to September 2015, the CI nowcasting lead time has a mean 27 value of ~64 minutes, with a longest lead time over 120 minutes. It is suggested that 28 false-alarm nowcasts that occur in the presence of capping inversion require further 29 investigation and algorithm enhancements.

atmospheric condition and orographic forcing increase the probability of CI occurrence over 120 Fujian. Therefore, Fujian is chosen as the domain of interest in this study. angles. Based on this definition, 27 CI days with a total of 157 daytime and 59 nighttime CI 129 events are identified during the three-month period from July to September 2015 (Table 1). 130 The day and night discrimination is based on whether the solar zenith angle is less than 60° or 131 not. Among these 157 daytime CI events, only 35 did not undergo significant splitting or 132 merging, and they are used in the selection of CI predictors. The CI events occurring during 133 the winter season, although existing, are quite rare and not investigated in this study. 134 2.2 AHI data 135 The Japanese AHI/Himawari-8 has three visible channels, three near-infrared channels 136 and 10 infrared channels (Table 2). Compared to its predecessor, the Multi-functional 137 Transport Satellite (MTSAT) imagers with one visible and four infrared channels, the AHI 138 provides more channels to better characterize the surface and cloud top features, as well as 139 the vertical profiling of the atmosphere. The sub-satellite resolutions of AHI are 0.5 km for 140 channel 3, 1 km for channels 1, 2 and 4, and 2 km for the other 12 channels. In this study, the full-disk images of all 16 AHI channels are cropped to a proper size and then remapped into a 142 0.5-km Lambert Conformal projection, true at 30 o N and 60 o N. By doing so, the 143 low-resolution infrared channels can be directly combined with the visible channels, while 144 the fine detail provided by high-resolution visible channels can be preserved. The AHI cloud 145 type and atmospheric motion vector products are not employed in this study.

147
The CI predictors are selected from 12 channels or channel combinations (hereafter, 148 interest fields) defined by AHI visible and infrared channels. Although in principle the 16 149 AHI channels can be used to build a set of more than 100 interest fields by dual-or 150 triple-channel combinations, many have redundancy in physical attributions. For example, 151 measurements from channels 1-4 are sensitive to the cloud optical thickness and thus have a 152 correlation greater than 0.90 between any two of them. It is not necessary to use all the four 153 channels for constructing CI predictors reflecting cloud optical thickness. With regard to the 154 contrast between the surface and cloud top, channel 1 reflectance (ρ0.47) is the most 155 significant among channels 1-4 (Zhuge et al. 2017). Therefore, a subset of 12 candidate 156 interest fields are selected for describing the atmospheric states and cloud-top properties contents within upper, middle, and lower tropospheric layers, respectively. For 165 middle and high clouds, Tb,6.2,Tb,6.9,and Tb,7.3 remain unchanged unless the cloud top 166 reaches their peak weighting function levels. The BT difference (BTD) between 167 channels 9 and 10 (i.e., Tb,7.3) can be utilized for estimating whether a cloud has 168 grown to a given altitude (Matthee and Mecikalski, 2013). Tb,6.9-Tb,7.3 is mostly 169 negative (~ -15 K) when the moist layer is at ~500hPa and meanwhile the cloud top 170 is lower than ~600 hPa. An elevated cloud top or moist layer would increase the 171 value of Tb,7.3. 172 2) The BT of channel 13 (i.e., Tb,10.4) and the BTD between channels 10 and 13 (i.e., 173 Tb,10.4) are indicators of cloud-top height. As the cloud top is elevated, the value 174 of Tb, 10.4 decreases,while Tb,10.4 increases. Tb,10.4 will turn from negative to 175 zero values if the cloud top is elevated to upper troposphere (~300 hPa). According to 176 Walker et al. (2012), Tb,7.3-Tb,10.4 is also an indicator of mid-level capping inversion.

180
The difference between Tb,12.4 and Tb,10.4 is also determined by cloud-top effective 181 radius.

182
4) The tri-channel difference of channels 11,15 and 13 (i.e.,Tb,8.6+Tb,10.4) is 183 utilized to infer the cloud-top phase (Baum et al., 2000). The physical consideration 184 for cloud-top phase discrimination is based on the difference between ice and liquid water particle absorption spectra within the wavelength range 8-13 μm. The ice 186 absorption coefficient increases faster between 8 and 11 μm than between 11 and 12 187 μm, while the opposite is true for liquid water. As a result, ice (water) clouds tend to 188 have larger (smaller) values of Tb,10.4 than those of Tb,12.4. The tri-channel 189 difference is positive for ice-phase cloud top, while negative for water-phase cloud 190 top. An increasing tri-channel difference means the cloud-top is gradually glaciated.

191
Note that the tri-channel difference algorithm was found to be less accurate in regions 192 where more than one cloud phase types occur, and needs to be updated (Baum et al., 193 2012). 194 5) The reflectance values of channels 5 (i.e., ρ1.6), 6 (i.e., ρ2.3), and 7 (i.e., ρ3.9) are 195 correlated with the hydrometeor particle sizes (effective radius) in both liquid and ice 196 phases of water vapor (Nakajima and King, 1990). However, due to the differences of 197 the indices of refraction at different wavelengths (Mecikalski et al., 2010), 198 correlations between any two of ρ1.6, ρ2.3, and ρ3.9 are lower than 0.8. Therefore, these 199 three interest fields are all retained for the further assessment. ρ3.9 is derived from the

203
The CI nowcasting algorithm focuses mainly on immature cumulus clouds before they 204 grow into cumulonimbus clouds or precipitating thunderstorms. After a CI occurrence, a 205 thunderstorm would be formed and the CI detection is ''turned off'' by the CI nowcasting  Table 2. More details about the three components of the CI algorithm are described below.  Tb,12.4) at the target pixel is 0.6 K greater than the 231 BTD value at the warmest pixel within the 19×19 pixel box centered at the target pixel; or (iii) 232 the BTD between 8. Tb,12.4) is 1.6 K greater than the BTD value at 233 the warmest pixel within the 19×19 pixel box centered at the target pixel. The (ii) and (iii) 234 checks are used for detecting thin cirrus clouds, which were introduced by Krebs et al.

242
Obviously, the cumulus cloud mask algorithm is significantly improved over the 243 SATCAST algorithm (Walker et al., 2012). Firstly, the thick-cloud pixel identification is used 244 instead of the cloud type products so that the CI nowcasting algorithm is independent of 245 ancillary data. Secondly, the removal of thunderstorms is conducted after the cloud object 246 identification. By doing so, a misclassification of cirrus anvil and/or cloud edge of a 247 thunderstorm, which could have a BT value higher than -20°C, as an individual cumulus is 248 avoided. Assuming a wind speed of 5 m s -1 that is appropriate in most conditions, a cumulus cloud 260 would move to a position 3 km (or 6 pixels) away in 10 min with this speed. Since the 261 equivalent diameter of a cumulus object is larger than 3 km most of time, an overlap between 262 T1 and T2 objects would exist even though the T1 object is not advected.

264
The procedure of CI forecast determinations is similar to that in the SATCAST algorithm 265 except for the selected predictors and corresponding thresholds. The CI nowcastings are 266 provided using a binary, deterministic approach with several satellite-based CI predictors 267 associated with the cloud top properties. When a majority of the CI predictors meet their 268 thresholds, CI is likely to occur within the next 0-2 h and thus a positive CI forecast is 269 assigned for the cloud object, or else, a null forecast is assigned. To avoid the blurring of the interest fields listed in Table 3. The selection is based on the distributions of 12 interest fields 274 for the 35 CI events that did not undergo significant splitting or merging. The distributions 275 are different at different time prior to the CI occurrence. To help understanding the cumulus 276 evolution prior to CI occurrences, the distributions of the 12 interest fields during the period 277 from t0-60min to t0 at a 10-min interval will be shown below , where t0 is defined 278 as the most recent AHI scan time before a CI occurrence. In other words, if a CI occurred at 279 0136 UTC, then t0 is set to 0130 UTC. occurred in Fujian with a warm-rain process, since the cloud-top at t0 is still warmer than 0°C 283 (i.e., 273.15 K) for most CI events. It can also be inferred that the upper-and 284 mid-tropospheric water vapor contents did not undergo many changes prior to CI occurrences 285 ( Fig. 4e and 4f). The values of Tb,7.3 (Fig. 4g) and Tb,6.9-Tb,7.3 (Fig. 5c) began to change 20 min 286 prior to a CI occurrence, indicating that the cloud top is elevated to ~600hPa. The cloud 287 optical thickness indicated by both ρ0.47 (Fig. 4a) and Tb,10.4 (Fig. 5b), as well as the 288 cloud-top phase indicated by Tb,8.6+Tb,10.4 (Fig. 5d), also show an increasing trend, 289 which suggests that the cloud gradually deepened and the cloud top was gradually glaciated 290 before the CI occurrence. However, the particle size indicated by the three near-infrared or 291 mid-infrared interest fields, ρ1.6 ( Fig. 4b), ρ2.3 (Fig. 4c), and ρ3.9 (Fig. 4d), did not show a 292 consistent trend. The trends for ρ1.6 ( Fig. 4b) and ρ2.3 (Fig. 4c) are not monotonic, but the 293 trend for ρ3.9 (Fig. 4d) is. This inconsistence among the three near-infrared or mid-infrared 294 interest fields may be caused by differences in penetration depths of the three channels. The channel 7 (3.9μm) can only detect the upper portion of a convective cloud while the channel 296 5 (1.6μm) can detect the lower part. Inconsistency among the three channels may indicate 297 heterogeneous droplet size profile. In this study, ρ3.9, which indicates the cloud-top particle 298 size, is retained. Based on the above assessments, six interest fields-ρ0. 47,ρ3.9,Tb,10.4,299 Tb,10.4,Tb,10.4,and Tb,8.6+Tb,10.4, are finally selected as the CI predictors of 300 the proposed CI nowcasting algorithm. The thresholds for the six selected CI predictors are 301 determined based on the 25 th percentile 30 min prior to the CI occurrence. This ensures that 302 the average forecast lead time is about 30 min long.

303
The cloud updraft strength, deepening rate, as well as cloud-top hydrometeor growth and 304 glaciation rate monitored could be assessed by examining the 10-min temporal differences of 305 Tb, 10.4,ρ0.47,ρ3.9,and Tb,8.6+Tb,10.4 (Fig. 6). A weak updraft with a strength around 306 -1K/10min is found for most CI events (Fig. 6c). For most CI events, the updraft became 307 stronger 20 min prior to CI occurrences, and reached a peak value of -4K/10min when CI 308 occurs. Similarly, the particle size growth had been slow in the earlier stage and faster when 309 the cloud developed into a CI (Fig. 6b). Exceptions are found for about 25% of CI events to 310 experience negative updrafts and particle size growths. Since the rates of variations of both 311 the cloud optical thickness and cloud-top phase are not regular (Fig. 6a and 6d), only the 312 10-min temporal differences of Tb,10.4 and ρ3.9 are taken as the two CI predictors in addition to 313 the above six selected CI predictors (Table 4).

314
At the last step of CI forecast determination, a cumulus object will get a positive CI 315 forecast if a decreasing Tb,10.4 or ρ3.9 is observed during a 10-min time period and at least five 316 of the eight CI predictors need to meet their thresholds. Table 5 gives the PODs as functions as the number of CI predictors meeting their thresholds, where POD is defined as the fraction 318 of CI events that are correctly forecasted. To ensure the POD be greater than 75% at 30 min 319 prior to CI, at least five CI predictors seem to be an "optimal" choice. During nighttime, ρ0.47 320 and ρ3.9 are not available, and at least three of the five CI predictors meeting their thresholds 321 (Table 6) for a positive CI forecast. and developed into two severe multi-cell thunderstorms in the late afternoon (Fig. 7). We 328 may focus our attention on the time period from 0000 to 0300 UTC during which the earliest 329 15 CI events were triggered. Figure 8 presents six composite reflectivity radar images from 330 0200 UTC to 0300 UTC at a 12-min interval, along with the CIs determined by radar images.

331
The radar images at 0206, 0218, 0230, 0242 and 0254 UTC were not presented but the CIs 332 determined by radar images at these times are shown as dashed circles. As seen from these 333 radar images, two echoes with reflectivity intensities greater than 35 dBZ appeared at 0218 334 UTC, indicating the eruptions of the first two CI events. The number of intense echoes 335 increased rapidly with time. By 0300 UTC, 15 CI events were observed, which are marked 336 with solid or dashed circles on the radar images in Fig. 8f.

337
The AHI/Himawari-8 multispectral observations with a 10-min temporal resolution are 338 used to monitor the cumulus developments. To help understand how the CI algorithm works, two CI events, which are labeled as "A" and "B" in Fig. 8, are analyzed in more detail (Fig.   340   9). Both events took place at about 0220 UTC. The first prediction of event "A" is at 0050 341 UTC, preceding the actual occurrence by 88 min; while the first prediction of event "B" is at 342 0200 UTC, only preceding the actual occurrence by 24 min. The precursor of CI event "A", 343 hereafter cumulus "A", began with a relatively strong updraft around -2 K/10min. The 344 cloud-top temperature dropped below 10°C at 0040 UTC. After that, cumulus "A" developed 345 slowly, along with merging and splitting. A rapid growth of cumulus "A" was restored at 346 0150 UTC, with a strength about -5 K/10min. Since five or more CI predictors satisfy their 347 thresholds, a positive CI forecast is assigned for cumulus "A" beginning at time 0050 UTC.

348
In contrast, cumulus "B" developed slowly at the very beginning. The cloud top temperature 349 did not drop below 10°C until 0200 UTC. The cloud-top cooling rate was about -3 K/10min.

350
When the S-band CINRAD radar detected a CI signal, the cloud top of cumulus "B" was still 351 warmer than 0°C, indicating the associated precipitation resulted from a warm-rain process.

352
The CI forecasts determined by AHI/Himawari-8 from 0040 UTC to 0230 UTC on 1 353 August 2015 at a 10 min interval are shown in Fig. 10. The first positive CI forecast was 354 assigned on the images at 0040 UTC, and verified by the CI event that occurred at 0242 UTC.

355
The forecast lead time for this event is 122 min. Out of the 15 CI events at 0230 UTC, 13 356 were successfully predicted. The POD for this case is 93.33%. The CI event that is labeled as 357 "C" in Fig. 8 is not predicted by the CI nowcasting algorithm. We find that the cloud top 358 temperature of cumulus "C" remained warmer than 10°C even though the 35-dBZ echo was 359 detected by the S-band CINRAD radar. By examining the rain observations from the 360 rain-gauge stations and the sequence cloud development based on the AHI images (figure omitted), the CI signal of cumulus "C" might be inferred as a noise. The CI event "D" is 362 missed at 0230 UTC, but predicted at 0250 UTC. This is due to the fact that AHI did not 363 provide measurements at 0240 UTC. Otherwise, it could be predicted as early as 0240 UTC, 364 to give a 14-min forecast lead time. The AHI-based CI algorithm also produces seven false  from 14 min to more than 120 min. If the cumulus object continually develops with a rapid 382 rate, the lead time will be short. On the contrary, if the cumulus grows slowly after the cloud-top temperature dropped below 10°C, the early detection of CI will be relatively easy 384 and the lead time will be long. Another possible reason for a long lead-time of CI forecast 385 algorithm is the mountain slope. After a shallow cumulus cloud is initiated, the mountain 386 slopes may anchor the location of shallow cumulus that will deepen one hour or longer. A

394
Benefiting from high spatial-and temporal-resolution, and multi-spectral measurements 395 of the AHI instrument on board Himawari-8, the CI nowcasting algorithm is improved in the 396 following three aspects: 1) cumulus cloud mask, 2) cumulus-object tracking, and 3) CI 397 forecast determination. The AHI-based CI algorithm is independent of any dynamic ancillary 398 data, such as the cloud type and atmospheric motion vector products, and is able to process 399 the input data and generate CI forecasts immediately once the latest AHI data are received.

400
In order to make sure that the AHI-based CI algorithm is applicable to CIs in sub-tropical 401 regions, this study assessed 12 interest fields including the visible and near-infrared 402 reflectance, infrared water vapor and infrared window BTs, and dual-and triple-channel 403 BTDs. Finally, six interest field that are relevant to the cloud optical thickness, as well as 404 cloud-top height, partial size and glaciation, were selected to diagnose the developmental level of a certain cumulus. The temporal differences of cloud-top height and partial size are 406 selected to help diagnose whether a cumulus is growing. By using these eight CI predictors, 407 the AHI-based algorithm is skillful in predicting the CI occurrences associated with 408 warm-rain processes.

409
Validation of the AHI-based CI nowcasting algorithm was firstly performed with a case PODs are as high as 96.82% and 96.61% at daytime and nighttime, respectively. The forecast 417 lead time ranged from 14 to 120 min. The mean forecast lead time is ~64 min.

418
In the future, we will improve the AHI-based CI algorithm by further reducing false 419 alarms. The current version of the CI algorithm treats each CI predictor equally. False-alarm 420 forecasts might be generated when several relatively less sensitive predictors meet their 421 thresholds while others don't (Lee et al., 2017). To avoid this from happening, the relative 422 importance (weights) of different CI predictors will be computed by using the linear 423 discriminant analysis (Wilks, 2006). All CI predictors will be weighted and then summed to 424 give a set of probabilistic forecasts (i.e., 0-100%).   ,10.4 Brightness temperature difference between channels 10 and 13

Acknowledgments
Cloud-top height relative to lower-troposphere (~600 hPa) Tb,10.4 Brightness temperature difference between channels 15 and 13 Cloud optical thickness Tb,7.3 Brightness temperature difference between channels 9 and 10 Cloud-top height relative to mid-troposphere (~450 hPa) Tb,8.6+Tb,10.4 Tri-channel difference of Channels 11,15 and 13 Cloud-top phrase  8.6+Tb,10.4 >-6.0℃ all-day 10-min temporal difference of ρ3.9 <0.0% daytime 10-min temporal difference of Tb,10.4 <0.0K all-day     of CI events is indicated by colors and line styles according to the legend. The CI events 577 labeled for "A", "B", "C", and "D" will be discussed later.      of CI events is indicated by colors and line styles according to the legend. The CI events labeled for "A", "B", "C", and "D" will be discussed later. Fig. 9. Temporal variations of the CI interest fields (a) ρ0.47, (b) ρ3.9, (c) Tb, 10.4,(d) Tb,10.4,(e) Tb,10.4, and (f) tri-spectral difference for CI events A (red) and B (blue) indicated in Fig. 8. Red and blue dashed lines indicate the interval when the cloud objects were splitting or merging. The threshold for each predictor is shown with dashed black line.